Hyperspectral reflectance imaging (HRI) is an emerging clinical tool for characterizing spatial and temporal variations in blood perfusion and oxygenation for applications such as burn assessment, wound healing, retinal exams and intraoperative tissue viability assessment. Since clinical HRI-based oximeters often use near-infrared (NIR) light, NIR-enabled mobile phones may provide a useful platform for future point-of-care devices. Furthermore, quantitative NIR imaging on mobile phones may dramatically increase the availability and accessibility of medical diagnostics for low-resource settings. We have evaluated the potential for phone-based NIR oximetry imaging and elucidated factors affecting performance using devices from two different manufacturers, as well as a scientific CCD. A broadband light source and liquid crystal tunable filter were used for imaging at 10 nm bands from 650 to 1000 nm. Spectral sensitivity measurements indicated that mobile phones with standard NIR blocking filters had minimal response beyond 700 nm, whereas one modified phone showed sensitivity to 800 nm and another to 1000 nm. Red pixel channels showed the greatest sensitivity up to 800 nm, whereas all channels provided essentially equivalent sensitivity at longer wavelengths. Referencing of blood oxygenation levels was performed with a CO-oximeter. HRI measurements were performed using cuvettes filled with hemoglobin solutions of different oxygen saturation levels. Good agreement between absorbance spectra measured with mobile phone and a CCD cameras were seen for wavelengths below 900 nm. Saturation estimates showed root-mean-squared-errors of 5.2% and 4.5% for the CCD and phone, respectively. Overall, this work provides strong evidence of the potential for mobile phones to provide quantitative spectral imaging in the NIR for applications such as oximetry, and generates practical insights into factors that impact performance as well as test methods for performance assessment.